Annotating different but similar classes as same when building an pre-trained Object Detection model

Question

Foreknowledge: I'm building a project with yolov8. I need to detect sheep and goats in images. Pre-trained models don't have goat class.

Question: Goats and sheep look very similar. In most cases, even I have difficulty distinguishing them. Would labeling goats as sheep negatively affect the model? If it does have a negative effect, would I get a negative result if I train the pre-trained model myself by adding the goat class?

Answer 1

Labeling goats as sheep would negatively affect the model's performance, especially if there are significant differences in the features between these two classes. Mislabeling the classes can lead to confusion and incorrect predictions, which can impact the model's accuracy and reliability.And also there is a chance for overfitting